1 Package loading

Here we show a report with plots to visualize the output of CloneTracer for sample A.2 which contains nuclear SNVs and CNVs

# set global chunk settings
knitr::opts_chunk$set(echo = TRUE)
knitr::opts_chunk$set(message = FALSE)
knitr::opts_chunk$set(warning = FALSE)

library(tidyverse)
library(Seurat)
library(ComplexHeatmap)
library(reticulate)
pd <- import("pandas")
library(circlize)
library(DiagrammeRsvg)
library(DiagrammeR)
library(qpdf)
library(rsvg)
source("funct_clonal_analysis.R")

pat <- "A.2"

2 Clonal analysis

Trees selected by CloneTracer are stored in trees_clonetracer.pdf file. The title indicates the tree index which is necessary to identify the tree in the ELBO plot (see below). In this case there is only one tree selected

# load pickle object
pickle <- pd$read_pickle(paste0("../output/",pat,".pickle"))

# get tree with highest evidence
tree_list <- plot_trees(pickle, clone_cols = F)

# save trees as pdf 
export_trees(tree_list, h = 350, w = 200, outdir = paste0("plots_", pat))

xfun::embed_file("plots_A.2/trees_clonetracer.pdf")

Download trees_clonetracer.pdf

2.1 ELBO plots

2.1.1 All trees

Here I plot the ELBO for the last iteration of the heuristic search. It is clear that tree 2 has a highest evidence compared to the rest.


print_elbo(pickle, first_iter = 700, max_elbo = "53000")

2.2 Clonal hierarchy

sel_tree_ind <- "1"

# make tree
sel_tree <- plot_trees(pickle, tree = sel_tree_ind)

render_graph(sel_tree$tree) %>% export_svg() %>% charToRaw() %>% 
    rsvg_png("plots_A.2/selected_tree.png",
             width = 600, height = 950)

render_graph(sel_tree$tree)

2.3 Heatmaps

Here we show a heatmap with the VAF of the SNVs in single cells as well as the clonal probabilities inferred by CloneTracer. For CNVs the scaled ratio of counts in the region of interest over total counts is shown. Celltypes of interest can be added as an Annotation column. A seurat can be provided with celltypes as metadata column.

2.3.1 Clonal probabilities

We can show the clone posterior probabilities of each single cell. One can clearly see that cells with low healthy probability have low number of counts on chromosome 7, indicating that they carry a monosomy.

# load Seurat object
seurat <- subset(readRDS(url("https://figshare.com/ndownloader/files/36434613?private_link=717cb824ed6e13de2bc6")),
                 patient == pat) 

# add simplified celltype labels (only needed for this particular object)
seurat <- add_simple_ct(seurat)
seurat$ct <- ifelse(seurat$ct %in% c("T cells", "NK cells", "B cells"), "T & B cells", "Myeloid cells")

# make sure cell_barcodes are the same between seurat and pickle objects
seurat <- RenameCells(seurat, new.names = gsub("(^.+-1).+", "\\1", colnames(seurat)))

heat_clones <- make_heatmap(pickle, 
                        seurat = seurat, 
                        tree = "1",
                        pat = pat,
                        celltype = T, 
                        ct_column = "ct",
                        prob_type = "clones",
                        cnv_type = c("Monosomy", "Trisomy"), 
                        cnv_pos = c(2,3),
                        middle_point = 0.5,
                        clust_rows = T)

draw(heat_clones$plot, annotation_legend_list = heat_clones$legend, merge_legend = TRUE)

2.3.2 Cancer probability

It is also useful to visualize the cancer probabilities computed as 1-posterior probability of being healthy

heat_cancer <- make_heatmap(pickle, 
                        seurat = seurat, 
                        tree = "1",
                        pat = pat,
                        celltype = T, 
                        ct_column = "ct",
                        cnv_type = c("Monosomy", "Trisomy"), 
                        cnv_pos = c(2,3),
                        prob_type = "cancer",
                        middle_point = 0.5,
                        clust_rows = T)

draw(heat_cancer$plot, annotation_legend_list = heat_cancer$legend, merge_legend = TRUE)

2.3.3 both

We can also plot both the clonal probabilities and the cancer posterior for each single cell on the same heatmap

heat_both <- make_heatmap(pickle, 
                        seurat = seurat, 
                        tree = "1",
                        pat = pat,
                        celltype = T, 
                        ct_column = "ct",
                        prob_type = "both",
                        cnv_type = c("Monosomy", "Trisomy"), 
                        cnv_pos = c(2,3),
                        middle_point = 0.5,
                        clust_rows = T)

draw(heat_both$plot, annotation_legend_list = heat_both$legend, merge_legend = TRUE)

2.4 UMAPs

We can visualize the posterior probabilities in a dimensionality reduction plot (UMAP, t-SNE, PCA). It should be contained in a Seurat object

2.4.1 Cancer probabilities

umap_leuk <- umap_leuk(pickle, seurat, tree = sel_tree_ind, reduction = "umap")

umap_leuk

2.4.2 Clones

If we want to visualize the clonal assignments the easiest is to select a posterior probability threshold for the discrete assignments of cells to clones. In our manuscript we use 0.8.

umap_clones <- umap_clones(pickle, seurat, tree = "1", reduction = "umap", post_thr = 0.8)

umap_clones

3 Add metadata to Seurat

We can add the cancer and clonal probabilities as well as the discrete assignments as metadata columns to a Seurat object

# add clonal information to Seurat
seurat <- add_meta(seurat, pickle, tree = "1")